Proceedings of the Annual Meeting of the Cognitive Science Society

The study of attitudes in the social psychology literature displays a dearth of computational modeling efforts. The principal modeling approach has been artificial neural networks, typically in the form of simple recurrent networks. The most recent and influential work in this vein relies on Ising-like or Hopfield-like models, with a focus on network properties and parameters such as system temperature and their effects on the dynamics of attitude formation. This work, however, is seldom informed by or integrated with contemporary cognitive modeling. This affects (i) the broader validity of the social psychology approach, (ii) its ability to account for learning in a principled way, and (iii) an understanding of the dynamics of attitude retrieval. We describe two studies that provide a simple but direct comparison between the social psychology approach and cognitive modeling, focusing on characterizing the performance differences between the two modeling paradigms.